Time-Dependent Association Between Cardiotocographic Features and Hypoxic-Ischemic Encephalopathy

Figures

• 

  FIGURE 1.. Inclusion and exclusion criteria of the infants included in the study.

• 

  FIGURE 2.. FHR patterns of an epoch that started 1 hour before delivery and ended 40 minutes before delivery. The FHR belongs to a healthy infant. PeriCALM Patterns identified the FHR baseline (cyan), accelerations (green), and decelerations (magenta). The blank spaces correspond to gaps and uninterpretable FHR segments which were removed from analysis.

• 

  FIGURE 3.. Epoch availability for the three infant groups during the last 12 hours of labor. The solid lines indicate the proportion of infants from each group that had CTG data available at each time, the shaded area indicates the 95% confidence interval of this estimate.

• 

  FIGURE 4.. Time-varying features. Normalized mutual information (NMI) between FHR features that rejected the null hypothesis, and the time to delivery (TTD). The features were extracted from the FHR baseline (cyan), acceleration (green), deceleration (magenta), and contraction events. The range of acceptance of the null hypotheses is indicated in blue. See Table 5 for the names of each feature number.

• 

  FIGURE 5.. Features with a time-invariant association with outcome class. Normalized mutual information (NMI) between FHR features that rejected the null hypothesis, and the outcome of labor. The features were extracted from the FHR baseline (cyan), acceleration (green), deceleration (magenta), and contraction events. The range of acceptance of the null hypotheses is indicated in blue. See Table 5 for the names of each feature number.

• 

  FIGURE 6.. Features with a time-varying association with outcome class. Normalized mutual information (NMI) between FHR features that rejected the null hypothesis, and the outcome of labor conditioned on time to delivery (TTD). The features were extracted from the FHR baseline (cyan), acceleration (green), deceleration (magenta), and contraction events. The range of acceptance of the null hypotheses is indicated in blue. See Table 5 for the names of each feature number.

• 

  FIGURE 7.. A Type I feature. (A) Average number of late decelerations per epoch for the healthy and pathological (acidosis and HIE) groups (mean ± standard error). (B) The cumulative distribution of the feature during the last 12 hours of labor. (C) Comparison of the TI and TV KS statistic for the binary separability of the healthy and pathological (acidosis or HIE) groups (mean ± standard deviation). (D) Comparison of the feature value for maximum separability using the TI and TV approaches (mean ± standard deviation). All plots showing data as function of the time to delivery also indicate with a black circle those epochs for which the KS test showed statistically significant differences between groups.

• 

  FIGURE 8.. A Type II feature (A) Approximate entropy of the FHR in deceleration events (r = 0.3) for the healthy and pathological (acidosis or HIE) groups (mean ± standard error). (B) The cumulative distribution of the feature during the last 12 hours of labor. (C) Comparison of the TI and TV KS statistic for the binary separability of the healthy and pathological groups (mean ± standard deviation). (D) Comparison of the feature value for maximum separability using the TI and TV approaches (mean ± standard deviation). (A-C) Indicate with a black circle those epochs for which the KS test showed statistically significant differences between groups.

• 

  FIGURE 9.. A Type III Feature. (A) Total length of deceleration events per epoch for the healthy and pathological (acidosis or HIE) groups (mean ± standard error). (B) The cumulative distribution of the feature during the last 12 hours of labor. (C) Comparison of the TI and TV KS statistic for the binary separability of the healthy and pathological groups (mean ± standard deviation). (D) Comparison of the feature value for maximum separability using the TI and TV approaches (mean ± standard deviation). All plots showing data as function of the time to delivery also indicate with a black circle those epochs for which the KS test showed statistically significant differences between groups.

• 

  FIGURE 10.. Comparison of the TI NMI with the TV NMI.

• 

  FIGURE 11.. (A) Total length of acceleration events per epoch for each of the healthy and pathological (acidosis and HIE) infant groups (mean ± standard error). (B) The cumulative distribution of the feature during the last 12 hours of labor. The green line showing the cumulative distribution of the feature for infants in the healthy group is hidden by the other curve which overlap with each other. (C) Comparison of the TI and TV KS statistic for the binary separability of the healthy and pathological groups (mean ± standard deviation). (D) Comparison of the feature value for maximum separability using the TI and TV approaches (mean ± standard deviation). All plots showing data as function of the time to delivery also indicate with a black circle those epochs for which the KS test showed statistically significant differences between groups.